A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Filters
Darker than Black-Box: Face Reconstruction from Similarity Queries
[article]
2021
arXiv
pre-print
Preserved Fulltext
Other Versions
We propose a novel approach that allows reconstructing the face querying only similarity scores of the black-box model. ...
Although some of these approaches work in a black-box setup using only face embeddings, usually, on the end-user side, only similarity scores are provided. ...
, KIREEV, UDOVICHENKO, PETIUSHKO: DARKER THAN BLACK-BOX
RAZZHIGAEV, KIREEV, UDOVICHENKO, PETIUSHKO: DARKER THAN BLACK-BOX
RAZZHIGAEV, KIREEV, UDOVICHENKO, PETIUSHKO: DARKER THAN BLACK-BOX ...
arXiv:2106.14290v2
fatcat:ucp6rbp5nzaxzoo5k5cf5xhfmm
A Survey of Black-Box Adversarial Attacks on Computer Vision Models
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Other Versions
Hence, in this paper, we propose to present a comprehensive comparative study of various black-box adversarial attacks and defense techniques. ...
significant amount of research has focused on the reliability of such models, primarily in two aspects - white-box, where the adversary has access to the targeted model and related parameters; and the black-box ...
It is described using synsets similar to WordNet. There are more than 20,000 categories that have been annotated using more than 100,000 synset categories from WordNet. ...
arXiv:1912.01667v3
fatcat:cq5kjjta2zhxfirrj2tevn7noi
A Survey of Privacy Attacks in Machine Learning
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2020 pre-print arXiv:2007.07646v1 fatcat:5lyae4ccsvd5rgzhua7llratzm
Our contribution in this research is an analysis of more than 40 papers related to privacy attacks against machine learning that have been published during the past seven years. ...
Although the body of work in privacy has been steadily growing over the past few years, research on the privacy aspects of machine learning has received less focus than the security aspects. ...
The only black-box reconstruction attack until now was proposed by Yang et al. [104] . ...
arXiv:2007.07646v2
fatcat:sj7z2h2dhfdybgc234ha6hkrdq
Black-Box Approaches to Fair Machine Learning
2021
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Black-box methods, which require only query accessto the model, are well suited for this purpose, as they are feasible to carry out without knowing the full details of the model. ...
These methods require only black-box access to the model, allowing them to be effective even when the model holder is uncooperative. ...
The implementation of a linear regression model can be reconstructed efficiently with black-box access by using Gaussian elimination. ...
doi:10.1184/r1/16920208
fatcat:ockrs6xjrfge5irodkrgdkwehq
Disentangled Neural Architecture Search
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
However, existing methods rely heavily on a black-box controller to search architectures, which suffers from the serious problem of lacking interpretability. ...
On the NASBench-101 dataset, DNAS achieves state-of-the-art performance of 94.21% using less than 1/13 computational cost of baseline methods. ...
than 1/13 computational cost (150 queries vs. 2000 queries). ...
arXiv:2009.13266v1
fatcat:uo4o26iwyvb23nmhhwbagxzx4u
DISSECT: Disentangled Simultaneous Explanations via Concept Traversals
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2022; you can also visit the original URL.
The file type is application/pdf.
Other Versions
By training a generative model from a classifier's signal, DISSECT offers a way to discover a classifier's inherent "notion" of distinct concepts automatically rather than rely on user-predefined concepts ...
several concepts, (2) are influential to a classifier's decision and are coupled to its reasoning due to joint training (3), are realistic, (4) preserve relevant information, and (5) are stable across similar ...
Explanations should produce the desired outcome from the black-box classifier f . ...
arXiv:2105.15164v4
fatcat:aymyl3j45rdcjisqy2oxld3yiy
Bayesian Optimization with Missing Inputs
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Bayesian optimization (BO) is an efficient method for optimizing expensive black-box functions. In real-world applications, BO often faces a major problem of missing values in inputs. ...
We first find a probability distribution of each missing value so that we can impute the missing value by drawing a sample from its distribution. ...
Given a historical data [X, y] and a black-box function f : X → R (X is the input domain), X may contain missing values as mentioned in Case 1, and if we query a point x t ∈ X to compute the function value ...
arXiv:2006.10948v1
fatcat:3hepgwxskfeullwnu3gbnlorbi
GRAPHITE: Generating Automatic Physical Examples for Machine-Learning Attacks on Computer Vision Systems
[article]
2022
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2021 pre-print arXiv:2002.07088v4 fatcat:i7nk4k2eercado642ee7p6ph5mOther Versions
GRAPHITE is also flexible as it can easily trade-off transform-robustness, perturbation size, and query count in black-box settings. ...
., robustness of a attack to environmental physical variations such as viewpoint and lighting changes, and 3) supporting attacks in not only white-box, but also black-box hard-label scenarios, so that ...
Conversely, darker areas in Fig. 4a are good candidates for elimination from the mask. ...
arXiv:2002.07088v6
fatcat:gbgooezwcfa7bduoupz6leg4du
Trustworthy AI: A Computational Perspective
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Other Versions
In this survey, we present a comprehensive survey of trustworthy AI from a computational perspective, to help readers understand the latest technologies for achieving trustworthy AI. ...
An adversary aims to extract the model information by querying the machine learning model in a black-box setting such that he can potentially fully reconstruct the model or create a substitute model that ...
In [55, 168] , the authors showed that face recognition systems work better for white compared to darker faces. ...
arXiv:2107.06641v3
fatcat:ymqaxvzsoncqrcosj5mxcvgsuy
Biometric, Fingervein, Machine L Implementation of a Security System, Using Captured Fingervein, Applying the Concepts of Machine Learning
2020
International Journal Of Engineering And Computer Science
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
This work also includes the implementation of an Android fingerprint biometric system using the Android Near InfraRed (NIR) module, which exists to show the similarities and differences between the two ...
However, a "black-box" classifier cannot be blindly trained which assumes that features are ordered. ...
These peaks correspond to the boundary of the finger and scanning lens, which contains the brightest pixels (since brighter pixels have higher intensity values than darker ones). ...
doi:10.18535/ijecs/v9i04.4470
fatcat:5ijoqsgbjzaynijhincaxi642q
Object Recognition: Physiological and Computational Insights
[chapter]
2010
Primate Neuroethology
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2017; you can also visit the original URL.
The file type is application/pdf.
(C) Reconstruction of the test shape from the population response in B. ...
The contrast between the inside and the outside of the circle was not critical (G), but the spots and the bar had to be darker than the disk (H). ...
doi:10.1093/acprof:oso/9780195326598.003.0024
fatcat:ry5wi3rfczdixcyxmktjbtfmn4
Emergent linguistic structure in artificial neural networks trained by self-supervision
2020
Proceedings of the National Academy of Sciences of the United States of America
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
Indeed, we show that a linear transformation of learned embeddings in these models captures parse tree distances to a surprising degree, allowing approximate reconstruction of the sentence tree structures ...
J.H. and C.D.M. were partly funded by a gift from Tencent Corp. ...
However, while this black box approach teaches us that the model learns some form of syntax, it does not tell us how. ...
doi:10.1073/pnas.1907367117
pmid:32493748
fatcat:qzfpgwccgfcc7joah43nurdmze
Deep Neural Network Perception Models and Robust Autonomous Driving Systems
[article]
2020
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2020; you can also visit the original URL.
The file type is application/pdf.
In a black-box attack the adversarial method does not have access to the model, so the adversary has to query the target model in order to estimate those needed aspects of the model's interior structure ...
ATTACKS ARE CHARACTERIZED BASED ON WHETHER THEY ARE WHITE-BOX OR BLACK-BOX (I.E., HAVING ACCESSING TO MODEL DETAILS OR NOT), THE NUMBER OF NETWORK QUERIES
Table VII VII summarizes the main tech-
niques ...
arXiv:2003.08756v1
fatcat:w2ajk6ryrzg4bl5avd3thod5ti
In Search of Lost Space: Stan Douglas's Archaeology of Cinematic Darkness
2012
October
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Douglas's black box galleries were never entirely black. ...
The darkness of a cinema or black box gallery establishes a paradigm very different from that offered by Minimalism. ...
That recognizability is preserved in the exhibition history of Overture-a history that traverses nascent black-box galleries and aging movie theaters. ...
doi:10.1162/octo_a_00084
fatcat:tjm2euunbva3rooziaxgghoxoy
Developing Future Human-Centered Smart Cities: Critical Analysis of Smart City Security, Interpretability, and Ethical Challenges
[article]
2021
arXiv
pre-print
Preserved Fulltext
A copy of this work was available on the public web and has been preserved in the Wayback Machine. The capture dates from 2021; you can also visit the original URL.
The file type is application/pdf.
Note that this fulltext copy is not of the "primary" version of this work. The version it corresponds to is:
2021 pre-print arXiv:2012.09110v2 fatcat:5ki3bibid5as7nq7ztma5lja64Other Versions
As its name implies, the ultimate objective of the adversary is to clone or reconstruct the target model, reengineering a black-box model, or to compromise the nature and the properties of the training ...
[138] Facial Expression Recognition Model Extraction
CNNs Multiple dataset including AR Face [139] , BU3DFE [140] , and JAFFE [141] It aims to analyze whether a black box CNNs model can be steal ...
arXiv:2012.09110v4
fatcat:yxh5tvpehbgldcblweoovbvdsq
« Previous
Showing results 1 — 15 out of 291 results